We will have the DUNE General computing meeting:
Michael Kirby is inviting you to a scheduled Zoom meeting.
Topic: General DUNE Computing meeting
Time: Mon Nov 4, 2019 09:30 CT
Join Zoom Meeting
https://fnal.zoom.us/s/6308404326
One tap mobile
+16699006833,,6308404326# US (San Jose)
+16465588656,,6308404326# US (New York)
Dial by your location
+1 669 900 6833 US (San Jose)
+1 646 558 8656 US (New York)
Meeting ID: 630 840 4326
Find your local number: https://zoom.us/u/acoyhG3WuF
dunetpc v08_36_00 is the most recent release, but the mac builds failed. The reason is understood and fixed, but we did not retroactively fix v08_36_00. protoduneana is only built and released on request at the moment. We could go to a more formal procedure. As it is, product_deps doesn't keep up with dunetpc's version unless people keep it up.
Tom ran valgrind's massif tool on a decoder job and also protoDUNE_SP_keepup_decoder_reco.fcl for two events. A bit of low-hanging fruit -- 88 MB in two diagnostic histograms in the SSP raw decoder have now been commented out -- they do not belong in the production workflow. Tthe main memory usage in the reco job is by recob::Wire (created by StandardAdcWireBuildingService::build), something produced by WireCell::SimpleTrace::SimpleTrace, temporaries created by TensorFlow, and ROOT's output buffering.
Tom and David have been discussing with Saba Sehrish, Erica Snider and Kyle Knoepfel about making our dataprep stage thread-safe. We have updated our prep chain and told them about our new toolchain. There are some opportunities for threading.